Synthesis and Anti‐Mycobacterium tuberculosis Activity of Imidazo[2,1‐b][1,3]oxazine Derivatives against Multidrug‐Resistant Strains

The emergence of multidrug‐resistant strains of M. tuberculosis has raised concerns due to the greater difficulties in patient treatment and higher mortality rates. Herein, we revisited the 2‐nitro‐6,7‐dihydro‐5H‐imidazo[2,1‐b][1,3]oxazine scaffold and identified potent new carbamate derivatives having MIC90 values of 0.18–1.63 μM against Mtb H37Rv. Compounds 47–49, 51–53, and 55 exhibited remarkable activity against a panel of clinical isolates, displaying MIC90 values below 0.5 μM. In Mtb‐infected macrophages, several compounds demonstrated a 1‐log greater reduction in mycobacterial burden than rifampicin and pretomanid. The compounds tested did not exhibit significant cytotoxicity against three cell lines or any toxicity to Galleria mellonella. Furthermore, the imidazo[2,1‐b][1,3]oxazine derivatives did not show substantial activity against other bacteria or fungi. Finally, molecular docking studies revealed that the new compounds could interact with the deazaflavin‐dependent nitroreductase (Ddn) in a similar manner to pretomanid. Collectively, our findings highlight the chemical universe of imidazo[2,1‐b][1,3]oxazines and their promising potential against MDR‐TB.


Introduction
Even~11,000 years after the first known case of tuberculosis (TB), the disease remains one of the greatest threats to global public health. [1,2] TB is the second leading cause of death from an infectious agent, behind only COVID-19. [3] According to the latest report by the World Health Organization (WHO), TB was responsible for the deaths of 1.4 million people worldwide in 2021. [3] In comparison, COVID-19 was responsible for the deaths of 3.5 million people in the same year. [4] Among these deaths from TB, 191,000 were caused by resistant strains of Mycobacterium tuberculosis (Mtb). [3] Drug-resistant forms of the disease (DR-TB) are currently the main challenge for health authorities and the scientific community in the fight against TB. [5][6][7] Currently, there are 3 classifications for DR-TB: i) multidrug or rifampicin (RIF)-resistant TB (MDR/RR-TB), classified as resistant to at least RIF and isoniazid (INH) or only RIF; ii) pre-extensively drug-resistant TB (pre-XDR-TB), defined as MDR-TB plus additional resistance to any fluoroquinolone; and iii) extensively drug-resistant TB (XDR-TB), now reclassified as MDR-TB plus additional resistance to any fluoroquinolone and at least one more group A drug. [5,8] Drugs commonly used in the treatment of DR-TB include levofloxacin, bedaquiline, linezolid, clofazimine, delamanid, and amikacin. Many of these drugs are highly toxic and require a long period of treatment, which can exceed 18 months for longer MDR-TB regimens. This immense challenge highlights the need to continuously develop new therapeutic alternatives against DR-TB to achieve the 'End TB Strategy' goals. [3,9] Although the global scenario sounds alarming, several advances have been achieved in the development of new drugs against DR-TB. In recent years, regulatory agencies around the world have approved three new drugs after decades of inactivity in the discovery and development of new drugs against TB. Delamanid (1) and pretomanid (2) (Figure 1) were approved in 2014 and 2019, respectively. [10,11] These drugs are representative of the nitroimidazole class. Both are prodrugs that are activated by a deazaflavin-dependent nitroreductase (Ddn), primarily leading to the inhibition of mycolic acid biosynthesis. [12,13] However, an additional mode of action has been reported, which involves the generation of reactive nitrogen species, including nitric oxide, which interferes with cellular respiration. [14][15][16] Delamanid (1) was previously recommended by the WHO for use in longer MDR-TB treatment regimens. [17] Recently, the WHO released some updated guidelines for the treatment of DR-TB. Their report recommends implementing a new drug combination to treat almost all forms of DR-TB. The new recommendation includes BPaLM (a combination of bedaquiline, pretomanid, linezolid, and moxifloxacin) or BPaL (bedaquiline, pretomanid, and linezolid). [18] Despite the recent and significant advances these new drugs bring to the treatment outcomes of DR-TB, resistant strains of Mtb have already been identified, [19][20][21] highlighting the importance of the discovery and development of new compounds.
Recently, our research group identified highly selective benzofuroxan derivatives with potent activity against MDR-TB strains and in vivo sterilizing antitubercular activity. [22,23] Compounds 3 and 4 ( Figure 1) showed highly promising results, with MIC 90 values below 0.5 μM against several MDR strains of clinical isolates of Mtb. These analogues present the benzofuroxan moiety in their structure, which was shown to be critical for the potent anti-Mtb activity since its replacement with other heterocyclics led to a considerable loss of activity. [23] Therefore, in a continuing effort to develop new anti-TB molecules, we sought another heterocyclic system that could replace benzofuroxan. We selected the imidazo[2,1b] [1,3]oxazine moiety for this endeavour, which proved to be a highly beneficial choice. The 2-nitro-6,7-dihydro-5H-imidazo[2,1b] [1,3]oxazine subunit is the pharmacophore group in the antitubercular drug pretomanid (2) and is a close analogue of the 6-nitro-2,3-dihydroimidazo[2,1-b] [1,3]oxazole moiety present in delamanid (1). There is an immense amount of accumulated evidence related to the imidazo[2,1-b] [1,3]oxazine scaffold that places it in a prominent position in TB drug discovery. [24] Furthermore, a large number of imidazo [2,1b] [1,3]oxazine derivatives with potent and promising anti-TB activity have already been reported. [24][25][26][27][28][29][30] The drug design rationale behind the new series was based on previously reported studies of imidazo [2,1-b] [1,3]oxazines. For example, Blaser et al. [31] showed that changing the linker group from OCH 2 (2) to O-carbamate (5) (Figure 1) improved the in vitro potency by 4 times. Furthermore, when they switched to the arylpiperazine carbamate (6) (Figure 1), potency and solubility improved further compared to 5. [31] In another study, Rakesh et al. [32] reported a series of pretomanid-oxazolidinone hybrids as anti-TB agents. The introduction of a 3fluorophenyl attached to a cyclic amine (7) (Figure 1) was shown to lead to improved physicochemical properties compared to pretomanid (2). [32] However, unlike 6, none of the compounds tested in this latter series exhibited any significant efficacy in the chronic mouse model of Mtb infection. [32] Taking into account these data, the new series presented here was designed to incorporate the 2-nitro-6,7-dihydro-5H-imidazo [2,1b] [1,3]oxazine "warhead", O-carbamate as a linker, and the best 3-fluorophenyl side chains identified in our previous study. [23] This strategy led us to the identification of a new and more potent series of imidazo [2,1-b] [1,3]oxazine derivatives against MDR-TB. Additionally, most of the compounds were also capable of inhibiting intramacrophage mycobacterial growth and were not toxic in the different models evaluated.
The imidazo[2,1-b] [1,3]oxazine derivative 56, containing a sulphonamide in the structure, was the only analogue with a MIC 90 value greater than 1 μM (1.63 μM). The less potent activity of 56 may be related to the physicochemical properties of the molecule. Compound 56 had the lowest cLogD value (1.51) within the series. All other compounds displayed cLogD values greater than 2.7 and were more potent than compound 56. These data suggest the influence that physiochemical properties, especially logD, exert on the activity of this series. Intriguingly, several other sulphonamide-containing pretomanid derivatives were previously reported to have inferior activity, implying a poor tolerance for this group across the imidazo[2,1b] [1,3]oxazine class. [28,31] Although 56 has a higher MIC compared to the other analogues, it still has significant activity, being more potent than several anti-TB compounds currently in preclinical studies. [5] Despite the difficulty in establishing an SAR relationship, we can observe some patterns. For instance, compounds 54 and 55, with MIC 90 values of 0.18 μM, are very similar and the only difference between them is the substitution of the phenyl ring (54) for a pyridine ring (55). Additionally, both analogues contain a trifluoromethyl substituent in the structure, suggesting the importance of this group for antimycobacterial activity. Compounds 48, 49, and 50 also display similar structures, with the only difference being the cyclic ring attached to the fluorophenyl ring. It is evident here that compounds with higher cLogD values demonstrated more potent activity. For example, compounds 48 and 49, which bear a piperidine and a thiomorpholine ring, respectively, were more active than the morpholine analogue (50). Compound 50 has a cLogD of 2.72 while compounds 48 and 49 have higher cLogD values, 3.79  Cytotoxicity studies were carried out in three different cell lines: HepG2 (ATCC HB-8065; liver), MRC-5 (ATCC CCL-171; lung), and J774A-1 (ATCC TIB-67; murine macrophage) according to previously reported procedures. [40,41] The results were expressed as inhibitory concentrations (IC 50 ), and the selectivity indices (SI) were calculated as the ratio between the IC 50 and MIC 90 values. Importantly, none of the imidazo[2,1b] [1,3]oxazine derivatives showed cytotoxicity in the cell lines evaluated. The IC 50 values for all the analogues were greater than 322 μM in HepG2, 147 μM in MRC-5, and 106 μM in J774A-1, translating into high SI values for all compounds tested.

In vitro determination of activity against intramacrophage Mtb
During active infection and also in the dormant state, Mtb lodges mainly inside macrophages. [42,43] Therefore, compounds developed against TB must be able to cross not only the thick lipid wall of Mtb but also the cell membrane of macrophages to reach the main reservoir of mycobacteria. Evaluation of intramacrophage activity against Mtb is of significant importance in the initial stages of the development of new anti-TB drugs. [44] Here, we assessed the intramacrophage activity of the compounds in J774A.1, a mouse macrophage cell line, and the results are shown in Figure 2. All compounds were tested at their MIC 90 concentrations, as previously stipulated in the REMA assay, against Mtb H37Rv (ATCC 27294). The analogues showed significant inhibition of intracellular bacterial growth in a dosedependent manner compared to controls. For example, several compounds, including 47 and 51, showed a 2-log reduction in bacterial burden compared to the positive control. Furthermore, these compounds were able to inhibit bacterial growth much more effectively than the drugs used as controls, rifampicin and pretomanid. Although these drugs are currently used in the clinic, the new analogues described herein were superior in inhibiting intramacrophage mycobacterial growth. For example, compounds 47 and 51 reduced the bacterial load by more than 1 log relative to the rifampicin and pretomanid-treated infected macrophages.

Further in vitro activity against Mtb drug-resistant strains
The emergence of DR strains is the main concern of health authorities around the world. Currently, drug discovery campaigns aim to discover potential drugs against resistant forms of TB (MDR-and XDR-TB). [5] Consequently, we tested all of the compounds described here against a panel of clinical isolates (CI). These strains were phenotypically characterized and exhibited resistance to several anti-TB drugs. [45] Specifically, each CI strain exhibited resistance to at least the following drugs: CI-3: RIF; CI-4: INH and RIF; CI-5: INH and RIF; CI-6: INH and RIF; CI-7: INH and RIF. CI-1 and CI-2 were characterized as susceptible strains. The strains were classified as MDR based on the MIC 90 values of the control drugs against them. Table 2 shows the results of the in vitro evaluation against this panel of clinical isolates. The concentration for the determination of promising compounds versus resistant strains was defined as 5 μM. Overall, all the compounds tested exhibited potent activity against most of the clinical isolates assayed. For instance, the most promising compounds, 47, 48, 49, 51, 52, 53 and 55 displayed MIC 90 values below 0.5 μM against most clinical isolates, including MDR strains. As noted for the susceptible strain (H37Rv), the characterization of an SAR is a challenge, and it is not possible to establish a clear relationship between the antimycobacterial activity and the chemical structure of these compounds.

In vitro determination of activity against Gram-positive and negative bacteria and Candida albicans
Treatment of TB requires a long period. Hence, it is essential that the drugs used are highly selective against Mtb. Therefore, we tested the analogues against a panel of Gram-positive and negative bacteria and Candida albicans. Table 3 shows the results against 4 different bacterial species, namely Escherichia coli (ATCC 25922), Staphylococcus aureus (ATCC 25923), Salmonella typhi (ATCC 14028) and Pseudomonas aeruginosa (ATCC 27853), as well as the results against a fungus commonly found in the human microbiota, Candida albicans (ATCC 90028). As shown in Table 3, none of the compounds showed significant antibacterial or antifungal activity against the evaluated strains.

In vivo evaluation of acute toxicity in Galleria mellonella
After the initial screening, we decided to proceed with evaluation of the toxicity of the compounds in an in vivo model to ensure the safety of the compounds for future preclinical  evaluation. We selected Galleria mellonella larvae as a model for the acute toxicity test. [46] Compounds 47 and 49 were selected for this assay because they showed the best solubility profile in the lipid-soluble vehicles used in the assay. For the assay, compound 47 was solubilized at the ideal concentration of 2,000 mg/kg of mean body weight of the larvae group, while compound 49 was only able to be solubilized at half of this concentration (1,000 mg/kg of body weight). The evaluated compounds did not show any toxicity to the larvae at the concentrations tested ( Figure 3).

Molecular docking studies and in silico prediction of ADME properties
To comprehend the poses and interactions of the synthesized compounds (47-56) and the activating deazaflavin-dependent nitroreductase holoenzyme (h-Ddn) of Mtb, an in silico model was built, starting with the 3D structure of full-length h-Ddn generated from the AlphaFold web server (Supplementary Material: Figure S63). [47] The docking calculations were based on the modelled h-Ddn structure and the parent drug pretomanid.
Validation of the study was carried out by observing interactions involving the key amino acid residues F17, S78, and Y133, and the 2-nitro-6,7-dihydro-5H-imidazo[2,1-b] [1,3]oxazine moiety in its suggested orientation, according to the literature. [48][49][50][51] Compared to pretomanid, all synthesized compounds (47-56) ( Figure 4) showed similar interactions in the catalytic pocket of h-Ddn. Most of the compounds interact through hydrogen bonds involving the Y133 and S78 residues, at distances ranging from 2.1 Å (compound 56) to 2.7 Å (compound 55) and from 1.5 Å (compound 54) to 2.1 Å (compound 55), respectively. Herein, we found that the Y133 residue plays a key role in the catalytic activity of the enzyme by also forming a hydrogen bond with the required cofactor F420, as reported. [48,50] Specifically, we observed a bifurcated hydrogen bond [52] between the hydroxyl group of Y133, the C4 carbonyl oxygen of the F420 isoalloxazine moiety, and the nitro group of the imidazo[2,1b] [1,3]oxazine scaffold, which helps to anchor these two ring systems. Interestingly, the latter binding pattern (involving Y133 and the nitro group of pretomanid) was not identified in a previous molecular dynamics study based on a reconstructed homology model of full-length h-Ddn. [50] Additionally, our results revealed that the Y130 amino acid residue does not   interact with the ligand, contrary to the proposal made by Cellitti et al. (2012) (these authors modelled the nitroimidazooxazine core of pretomanid into a cocrystal structure of F420 bound to a truncated Ddn construct lacking 40 residues from the N-terminus). [48] Here, the afore-mentioned molecular dynamics studies conducted by Mohamed et al. (2016) support our finding, indicating that the role of the Y130 residue is to act as part of a hydrophobic barrier (together with Y65 and Y136) in the active site of Ddn. [50] Like the latter authors, we also observed π-π stacking interactions with cofactor F420 (cF420) for all compounds. This is in line with an activation mechanism that involves hydride transfer from C5 of the deprotonated cofactor to the C3 atom of the imidazo[2,1-b] [1,3]oxazine substrate (ultimately leading to the release of reactive nitrogen species). [50] Furthermore, the tail portion of all compounds studied showed intermolecular interactions with the α-helix of h-Ddn, particularly π-alkyl and/or π-π stacking interactions involving the residues F17 and W20, as well as sulfur-π or πalkyl interactions with M21, involving the side chain phenyl group. The best scores in Figure 4 affirm that interactions with aromatic residues belonging to the α-helix of h-Ddn are important for the activity, as mentioned by Cellitti et al. (2012), especially the interaction with W20. [48] Analysis of the docking results revealed that intermolecular interactions involving h-Ddn and the ligands 47, 49 and pretomanid occurred through the same amino acid residues.
The key amino acid residues Y133 and S78 interact with the mentioned ligands through hydrogen bond interactions with the nitro group of the imidazo[2,1-b] [1,3]oxazine moiety, as shown in Figure 5. Additionally, 47 forms a halogen bond between the fluorine atom in the aromatic subunit and the carbonyl group located in the backbone of F17 (Figure 5b). On the other hand, in Figure 5d, it was possible to identify the similar orientation of ligands 47 and 49; therefore, their similar biological activity could be explained by their interaction with the same regions of h-Ddn (catalytic pocket with hydrogen bond donor residues, cofactor F420 with π-alkyl and π-π stacking interactions and α-helix with aromatic hydrophobic residues). Although pretomanid also has the same orientation and interactions, rotatable covalent bonds at the final part of the pretomanid tail could allow additional interaction with complementary amino acid residues, such as F16 and W20. [48,50] Therefore, the conformational restrictions arising from terminal heterocyclic moieties in the tail of compounds 47 and 49 restrict the required torsion for interactions involving the F16 and W20 residues from the α-helix of h-Ddn.
In silico predictions of ADME properties were obtained by Swiss ADME software (Supplementary material). Compounds 47-56 were mostly soluble or moderately water-soluble and all have the potential to be absorbed after oral administration (> 82.6 %). The data suggest that compounds 47-56 are not substrates for CYP3A4, CYP2D6, and CYP1A2; however, all seem to be substrates for CYP2C19. In silico data also suggest that compounds 47-56 are not hERG inhibitors. Finally, compounds 47-49 did not violate any parameter of Lipinski's rules.

Conclusions
Herein, we present the continuation of our medicinal chemistry campaign that aimed to develop new anti-TB compounds. We designed and synthesized a series of new imidazo[2,1b] [1,3]oxazine derivatives through a lead optimization strategy based on benzofuroxan derivatives previously identified in our research group. The selection of the 2-nitro-6,7-dihydro-5Himidazo[2,1-b] [1,3]oxazine subunit to replace benzofuroxan was motivated by the fact that it is the pharmacophoric group of the anti-TB drug pretomanid and compounds containing this scaffold often exhibit high selectivity and potency against Mtb. Furthermore, we have included the O-carbamate group as a linker and a series of different side chains containing the 3fluorophenyl attached to a cyclic amine. In total, we synthesized and characterized ten new imidazo [2,1-b] [1,3]oxazine derivatives. We tested these compounds in a series of experiments designed to assess their potential as anti-TB agents, including evaluation against susceptible strains and MDR clinical isolates of Mtb, selectivity against a panel of different microorganisms, activity against intramacrophage Mtb, and toxicity toward in vitro and in vivo models. The new series of imidazo [2,1b] [1,3]oxazine derivatives showed high potencies against Mtb H37Rv, with MIC 90 values ranging from 0.18 to 1.63 μM. Additionally, most of the compounds displayed MIC 90 values below 0.5 μM against a panel of MDR clinical isolates. Finally, the compounds did not show toxicity toward in vitro or in vivo models and showed potent intramacrophage activity. Molecular modelling studies suggest that all compounds interact with the deazaflavin-dependent nitroreductase holoenzyme in poses similar to those found for pretomanid. The results presented here demonstrate the potential of this new series of compounds. The compounds showed potent activity against several MDR strains of Mtb and thus will advance to the next stages of development in due course.

Experimental Section Chemistry
All reagents, chemicals, and solvents were purchased from commercial suppliers and used without further purification. Reactions were monitored by thin-layer chromatography (TLC) using commercially available precoated plates and visualized with UV light at 254 nm. Flash column chromatography was carried out using Sigma Aldrich silica gel (pore size 60 Å, particle size 35-75 μm). The 1 H NMR and 13 C NMR spectra were obtained with a Bruker® Avance 400 MHz spectrometer at room temperature (rt) using deuterated chloroform (CDCl 3 ) or dimethylsulfoxide (DMSOd 6 ) as solvents. Chemical shifts were expressed in parts-per-million (ppm) relative to tetramethylsilane (TMS). The coupling constants (J) are denoted in hertz (Hz). Spin multiplicities were reported as singlet (s), doublet (d), triplet (t), quartet (q), multiplet (m), doublet of doublet (dd) and doublet of doublet of doublets (ddd). HPLC purity analysis for the final compounds confirmed that they were greater than 95 %. HPLC analysis was carried out using a Shimadzu® HPLC model CBM20-A coupled with a UV-VIS detector (model SPD-

General method for carbamate synthesis (47-56)
Compound 15 (26 mg, 0.144 mmol) was suspended in 3 mL of dry THF (3 mL) under N 2 (sealed after degassing) and treated with Et 3 N (90 μl, 0.575 mmol), then cooled on ice. Triphosgene (51 mg, 0.173 mmol) was added, and the mixture was sealed under N 2 and stirred at 20°C for 2 h (a white suspension was formed). A solution of the appropriate amine (37-46) (0.144 mmol) in dry THF (1 mL) was added using a syringe and the resulting mixture was stirred at 20°C for 14 h. The reaction mixture was quenched with aqueous NaHCO 3 , then water was added, and the mixture was extracted with CH 2 Cl 2 . The combined extracts were dried over Na 2 SO 4 and then evaporated under reduced pressure. The crude product was purified by flash chromatography using EtOAc:hexane (90 : 10, v/v) as the mobile phase to give the final carbamates (47-56) as offwhite solids in varying yields.